Policy-based optimization for matching validation algorithm in monocular robotics

A novel policy to use the HOHCT validation algorithm is presented. HOHCT was introduced as a batch gating technique to validate data association in delayed initialization monocular SLAM. This new policy does not alter the inherent complexity of the algorithm, which lies in the exponential cost for worst case, but helps to keep it down to the average case-linear or quadratic-, while at the same time reducing the incidence of other operations with polynomial costs-obtaining new features for the map-. Statistics from off-line experiments with real data is use to evaluate costs and impact of the proposed policy.